Vowel Acoustics as Predictors of Speech Intelligibility in Dysarthria

This is the code for the statistical analysis for “Vowel Acoustics as Predictors of Speech Intelligibility in Dysarthria.”

Loading Packages

library(rio)

The following rio suggested packages are not installed: ‘csvy’, ‘feather’, ‘fst’, ‘hexView’, ‘readODS’, ‘rmatio’
Use 'install_formats()' to install them

library(tidyverse)

Registered S3 methods overwritten by 'dbplyr':
  method         from
  print.tbl_lazy     
  print.tbl_sql      
── Attaching packages ────────────────────────────────────────────────────────── tidyverse 1.3.0 ──
✓ ggplot2 3.3.5     ✓ purrr   0.3.4
✓ tibble  3.1.6     ✓ dplyr   1.0.7
✓ tidyr   1.1.3     ✓ stringr 1.4.0
✓ readr   1.4.0     ✓ forcats 0.5.1
── Conflicts ───────────────────────────────────────────────────────────── tidyverse_conflicts() ──
x dplyr::filter() masks stats::filter()
x dplyr::lag()    masks stats::lag()

library(irr) # install.packages('irr')

Loading required package: lpSolve

library(performance)

Attaching package: ‘performance’

The following object is masked from ‘package:irr’:

    icc

library(car)

Loading required package: carData
Registered S3 methods overwritten by 'car':
  method                          from
  influence.merMod                lme4
  cooks.distance.influence.merMod lme4
  dfbeta.influence.merMod         lme4
  dfbetas.influence.merMod        lme4

Attaching package: ‘car’

The following object is masked from ‘package:dplyr’:

    recode

The following object is masked from ‘package:purrr’:

    some

library(ggpubr)
library("Hmisc") # install.packages('Hmisc')

Loading required package: lattice
Loading required package: survival
Loading required package: Formula

Attaching package: ‘Hmisc’

The following objects are masked from ‘package:dplyr’:

    src, summarize

The following objects are masked from ‘package:base’:

    format.pval, units

library(ggridges)
library(furniture) # install.packages('furniture')
library(gt)

Attaching package: ‘gt’

The following object is masked from ‘package:Hmisc’:

    html

library(patchwork)
library(ks)
library(emuR) # install.packages('emuR')

Attaching package: ‘emuR’

The following object is masked from ‘package:Hmisc’:

    label

The following object is masked from ‘package:car’:

    ellipse

The following object is masked from ‘package:base’:

    norm

library(geomtextpath) # remotes::install_github("AllanCameron/geomtextpath")

Upload Datasets

Reliability <- rio::import("Prepped Data/Reliability Data.csv")
AcousticData <- rio::import("Prepped Data/AcousticMeasures.csv") %>%
  dplyr::mutate(intDiff = VAS - transAcc)

AcousticData <- AcousticData %>%
  dplyr::filter(!grepl("_rel", Speaker)) %>%
  dplyr::select(c(Speaker, Sex, Etiology, vowel_ED_b, VSA_b, Hull_b, Hull_bVSD_25, Hull_bVSD_50, Hull_bVSD_75, VAS, transAcc)) %>%
  dplyr::mutate(Etiology = as.factor(Etiology),
                Sex = as.factor(Sex),
                Speaker = as.factor(Speaker))

Listeners <- rio::import("Prepped Data/Listener_Demographics.csv") %>%
  dplyr::select(!c(StartDate:proloficID, Q2.4_6_TEXT, Q3.2_8_TEXT, AudioCheck:EP3))

Listeners$race[Listeners$Q3.3_7_TEXT == "Native American/ African amercing"] <- "Biracial or Multiracial"

Inter-rater Reliability

Two raters (the first two authors) completed vowel segmentation for the speakers. To calculate inter-rater reliability, 20% of the speakers were segmented again by the other rater. Two-way intraclass coefficients were computed for the extracted F1 and F2 from the temporal midpoint of the vowel segments. Since only one set of ratings will be used in the data analysis, we focus on the single ICC results and interpretation. However, we also report the average ICC values to be comprehensive.

## Creating new data frames to calculate ICC for extracted F1 and F2 values

F1_Rel <- Reliability %>%
  dplyr::select(c(F1, F1_rel))

F2_Rel <- Reliability %>%
  dplyr::select(c(F2, F2_rel))
  
## Single ICC for F1
Single_F1 <- irr::icc(F1_Rel, model = "twoway", type = "agreement", unit = "single")

## Average ICC for F1
Average_F1 <- irr::icc(F1_Rel, model = "twoway", type = "agreement", unit = "average")

## Single ICC for F2
Single_F2 <- irr::icc(F2_Rel, model = "twoway", type = "agreement", unit = "single")

## Average ICC for F2
Average_F2 <- irr::icc(F2_Rel, model = "twoway", type = "agreement", unit = "average")

## Inter-rater reliability results and interpretation

print(paste("Single ICC for F1 is ", round(Single_F1$value, digits = 3), ". ", 
            "The 95% CI is [", round(Single_F1$lbound, digits = 3), " - ", round(Single_F1$ubound, digits = 3), "].", sep = ""))

[1] "Single ICC for F1 is 0.866. The 95% CI is [0.837 - 0.89]."

print(paste("Single ICC for F2 is ", round(Single_F2$value, digits = 3), ". ", 
            "The 95% CI is [", round(Single_F2$lbound, digits = 3), " - ", round(Single_F2$ubound, digits = 3), "].", sep = ""))

[1] "Single ICC for F2 is 0.931. The 95% CI is [0.916 - 0.944]."

print(paste("Average ICC for F1 is ", round(Average_F1$value, digits = 3), ". ", 
            "The 95% CI is [", round(Average_F1$lbound, digits = 3), " - ", round(Average_F1$ubound, digits = 3), "].", sep = ""))

[1] "Average ICC for F1 is 0.928. The 95% CI is [0.911 - 0.942]."

print(paste("Average ICC for F2 is ", round(Average_F2$value, digits = 3), ". ", 
            "The 95% CI is [", round(Average_F2$lbound, digits = 3), " - ", round(Average_F2$ubound, digits = 3), "].", sep = ""))

[1] "Average ICC for F2 is 0.964. The 95% CI is [0.956 - 0.971]."

print("Thus, interrater reliability for the extracted F1 and F2 values from the vowel segments was good to excellent.")

[1] "Thus, interrater reliability for the extracted F1 and F2 values from the vowel segments was good to excellent."

## Removing extra data frames from environment

rm(F1_Rel, F2_Rel, Reliability, Single_F1, Single_F2, Average_F1, Average_F2)

Descriptive Statistics

Correlations

CorrMatrix <- AcousticData %>%
  dplyr::select(VSA_b, vowel_ED_b, Hull_b, Hull_bVSD_25, Hull_bVSD_50, Hull_bVSD_75, VAS, transAcc) %>%
  as.matrix() %>%
  Hmisc::rcorr()

CorrMatrix <- CorrMatrix$r

stats::cor.test(AcousticData$VSA_b, AcousticData$vowel_ED_b, method = "pearson")

    Pearson's product-moment correlation

data:  AcousticData$VSA_b and AcousticData$vowel_ED_b
t = 6.5285, df = 38, p-value = 1.076e-07
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
 0.5372669 0.8468014
sample estimates:
      cor 
0.7270881 

stats::cor.test(AcousticData$Hull_b, AcousticData$Hull_bVSD_25, method = "pearson")

    Pearson's product-moment correlation

data:  AcousticData$Hull_b and AcousticData$Hull_bVSD_25
t = 9.4906, df = 38, p-value = 1.43e-11
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
 0.7135136 0.9119080
sample estimates:
     cor 
0.838625 

write.csv(CorrMatrix, file = "Tables/Correlation Matrix.csv")
rm(CorrMatrix)

Research Q1: Modeling Intelligibility

Orthographic Transcriptions

Model 1

# Specifying Model 1

OT_Model1 <- lm(transAcc ~ Hull_bVSD_25, data = AcousticData)

## Model 1 Assumptions 

performance::check_model(OT_Model1)

## Model 1 Summary

summary(OT_Model1)

Call:
lm(formula = transAcc ~ Hull_bVSD_25, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-47.896 -14.090   5.996  17.470  36.105 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept)   48.2973     9.7372   4.960  1.5e-05 ***
Hull_bVSD_25   0.6417     0.5663   1.133    0.264    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 23.83 on 38 degrees of freedom
Multiple R-squared:  0.03268,   Adjusted R-squared:  0.007224 
F-statistic: 1.284 on 1 and 38 DF,  p-value: 0.2643

Model 2

## Specifying Model 2

OT_Model2 <- lm(transAcc ~ Hull_bVSD_25 + Hull_bVSD_75, data = AcousticData)

## Model 2 Assumption Check

performance::check_model(OT_Model2)

## Model 2 Summary

summary(OT_Model2)

Call:
lm(formula = transAcc ~ Hull_bVSD_25 + Hull_bVSD_75, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-48.854 -13.962   5.567  16.758  36.257 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept)   47.3374    10.3316   4.582 5.09e-05 ***
Hull_bVSD_25   0.8045     0.7781   1.034    0.308    
Hull_bVSD_75  -0.7188     2.3225  -0.309    0.759    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 24.11 on 37 degrees of freedom
Multiple R-squared:  0.03518,   Adjusted R-squared:  -0.01698 
F-statistic: 0.6745 on 2 and 37 DF,  p-value: 0.5156

## Model 1 and Model 2 Comparison

anova(OT_Model1, OT_Model2)

Analysis of Variance Table

Model 1: transAcc ~ Hull_bVSD_25
Model 2: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75
  Res.Df   RSS Df Sum of Sq      F Pr(>F)
1     38 21570                           
2     37 21514  1    55.696 0.0958 0.7587

Model 3

## Specifying Model 3

OT_Model3 <- lm(transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b, data = AcousticData)

## Model 3 Assumption Check

performance::check_model(OT_Model3)

## Model 3 Summary

summary(OT_Model3)

Call:
lm(formula = transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b, 
    data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-56.706 -13.157   7.018  17.957  29.990 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)  
(Intercept)   31.0806    14.4916   2.145   0.0388 *
Hull_bVSD_25  -0.8439     1.2984  -0.650   0.5199  
Hull_bVSD_75   0.3159     2.3714   0.133   0.8948  
Hull_b         1.2941     0.8247   1.569   0.1253  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 23.65 on 36 degrees of freedom
Multiple R-squared:  0.09695,   Adjusted R-squared:  0.0217 
F-statistic: 1.288 on 3 and 36 DF,  p-value: 0.2932

## Model 2 and Model 3 Comparison

anova(OT_Model2, OT_Model3)

Analysis of Variance Table

Model 1: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75
Model 2: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b
  Res.Df   RSS Df Sum of Sq      F Pr(>F)
1     37 21514                           
2     36 20137  1    1377.5 2.4626 0.1253

Model 4

## Specifying Model 4

OT_Model4 <- lm(transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b, data = AcousticData)

## Model 4 Assumption Check

performance::check_model(OT_Model4)

## Model 4 Summary

summary(OT_Model4)

Call:
lm(formula = transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + 
    VSA_b, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-44.796 -11.144   3.042  12.567  34.297 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)   
(Intercept)   27.4533    13.1156   2.093  0.04365 * 
Hull_bVSD_25  -1.2602     1.1782  -1.070  0.29214   
Hull_bVSD_75   0.5663     2.1390   0.265  0.79274   
Hull_b         0.7364     0.7654   0.962  0.34261   
VSA_b          6.0896     1.9954   3.052  0.00432 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 21.32 on 35 degrees of freedom
Multiple R-squared:  0.2868,    Adjusted R-squared:  0.2052 
F-statistic: 3.518 on 4 and 35 DF,  p-value: 0.01626

## Model 3 and Model 4 Comparison

anova(OT_Model3, OT_Model4)

Analysis of Variance Table

Model 1: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b
Model 2: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b
  Res.Df   RSS Df Sum of Sq     F   Pr(>F)   
1     36 20137                               
2     35 15904  1    4232.4 9.314 0.004321 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Model 5

## Specifying Model 5

OT_Model5 <- lm(transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b + vowel_ED_b, data = AcousticData)

## Model 4 Assumption Check

performance::check_model(OT_Model5)

## Model 4 Summary

summary(OT_Model5)

Call:
lm(formula = transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + 
    VSA_b + vowel_ED_b, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-46.533 -11.028   3.327  13.017  33.227 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)  
(Intercept)   20.3598    21.5072   0.947   0.3505  
Hull_bVSD_25  -1.2567     1.1924  -1.054   0.2994  
Hull_bVSD_75   0.7007     2.1882   0.320   0.7508  
Hull_b         0.6895     0.7826   0.881   0.3845  
VSA_b          5.3903     2.6194   2.058   0.0473 *
vowel_ED_b     5.5182    13.1650   0.419   0.6777  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 21.57 on 34 degrees of freedom
Multiple R-squared:  0.2904,    Adjusted R-squared:  0.1861 
F-statistic: 2.783 on 5 and 34 DF,  p-value: 0.03271

## Model 3 and Model 4 Comparison

anova(OT_Model4, OT_Model5)

Analysis of Variance Table

Model 1: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b
Model 2: transAcc ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b + vowel_ED_b
  Res.Df   RSS Df Sum of Sq      F Pr(>F)
1     35 15904                           
2     34 15823  1    81.764 0.1757 0.6777

Final Model

## Specifying Final Model

OT_Model_final <- lm(transAcc ~ VSA_b, data = AcousticData)

## Final Model Assumption Check

performance::check_model(OT_Model_final)

## Final Model Summary

summary(OT_Model_final)

Call:
lm(formula = transAcc ~ VSA_b, data = AcousticData)

Residuals:
   Min     1Q Median     3Q    Max 
-46.72 -12.69   2.97  14.37  35.39 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)   32.508      7.857   4.138 0.000187 ***
VSA_b          5.872      1.613   3.641 0.000807 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 20.86 on 38 degrees of freedom
Multiple R-squared:  0.2586,    Adjusted R-squared:  0.2391 
F-statistic: 13.25 on 1 and 38 DF,  p-value: 0.0008068

confint(OT_Model_final)

                2.5 %    97.5 %
(Intercept) 16.602765 48.413532
VSA_b        2.606927  9.137097

VAS Models

Model 1

# Specifying Model 1

VAS_Model1 <- lm(VAS ~ Hull_bVSD_25, data = AcousticData)

## Model 1 Assumptions 

performance::check_model(VAS_Model1)

## Model 1 Summary

summary(VAS_Model1)

Call:
lm(formula = VAS ~ Hull_bVSD_25, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-47.625 -16.684   8.462  19.440  37.352 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept)   42.6328    10.7512   3.965 0.000313 ***
Hull_bVSD_25   0.5877     0.6253   0.940 0.353236    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 26.31 on 38 degrees of freedom
Multiple R-squared:  0.02272,   Adjusted R-squared:  -0.003001 
F-statistic: 0.8833 on 1 and 38 DF,  p-value: 0.3532

Model 2

## Specifying Model 2

VAS_Model2 <- lm(VAS ~ Hull_bVSD_25 + Hull_bVSD_75, data = AcousticData)

## Model 2 Assumption Check

performance::check_model(VAS_Model2)

## Model 2 Summary

summary(VAS_Model2)

Call:
lm(formula = VAS ~ Hull_bVSD_25 + Hull_bVSD_75, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-47.850 -16.576   8.382  19.448  37.237 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept)   42.3384    11.4211   3.707 0.000684 ***
Hull_bVSD_25   0.6376     0.8602   0.741 0.463195    
Hull_bVSD_75  -0.2204     2.5674  -0.086 0.932036    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 26.66 on 37 degrees of freedom
Multiple R-squared:  0.02291,   Adjusted R-squared:  -0.0299 
F-statistic: 0.4338 on 2 and 37 DF,  p-value: 0.6513

## Model 1 and Model 2 Comparison

anova(VAS_Model1, VAS_Model2)

Analysis of Variance Table

Model 1: VAS ~ Hull_bVSD_25
Model 2: VAS ~ Hull_bVSD_25 + Hull_bVSD_75
  Res.Df   RSS Df Sum of Sq      F Pr(>F)
1     38 26296                           
2     37 26291  1     5.239 0.0074  0.932

Model 3

## Specifying Model 3

VAS_Model3 <- lm(VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b, data = AcousticData)

## Model 3 Assumption Check

performance::check_model(VAS_Model3)

## Model 3 Summary

summary(VAS_Model3)

Call:
lm(formula = VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-56.444 -18.989   6.121  18.036  32.281 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)
(Intercept)   25.0400    16.0599   1.559    0.128
Hull_bVSD_25  -1.1164     1.4389  -0.776    0.443
Hull_bVSD_75   0.8805     2.6280   0.335    0.740
Hull_b         1.3770     0.9139   1.507    0.141

Residual standard error: 26.21 on 36 degrees of freedom
Multiple R-squared:  0.08087,   Adjusted R-squared:  0.00428 
F-statistic: 1.056 on 3 and 36 DF,  p-value: 0.3799

## Model 2 and Model 3 Comparison

anova(VAS_Model2, VAS_Model3)

Analysis of Variance Table

Model 1: VAS ~ Hull_bVSD_25 + Hull_bVSD_75
Model 2: VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b
  Res.Df   RSS Df Sum of Sq      F Pr(>F)
1     37 26291                           
2     36 24731  1    1559.6 2.2702 0.1406

Model 4

## Specifying Model 4

VAS_Model4 <- lm(VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b, data = AcousticData)

## Model 4 Assumption Check

performance::check_model(VAS_Model4)

## Model 4 Summary

summary(VAS_Model4)

Call:
lm(formula = VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b, 
    data = AcousticData)

Residuals:
   Min     1Q Median     3Q    Max 
-42.25 -14.17   5.76  16.26  41.48 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)   
(Intercept)   21.0802    14.5922   1.445  0.15746   
Hull_bVSD_25  -1.5708     1.3109  -1.198  0.23885   
Hull_bVSD_75   1.1539     2.3798   0.485  0.63078   
Hull_b         0.7682     0.8516   0.902  0.37320   
VSA_b          6.6479     2.2200   2.995  0.00502 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 23.72 on 35 degrees of freedom
Multiple R-squared:  0.2683,    Adjusted R-squared:  0.1847 
F-statistic: 3.209 on 4 and 35 DF,  p-value: 0.02405

## Model 3 and Model 4 Comparison

anova(VAS_Model3, VAS_Model4)

Analysis of Variance Table

Model 1: VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b
Model 2: VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b
  Res.Df   RSS Df Sum of Sq      F  Pr(>F)   
1     36 24731                               
2     35 19687  1    5043.9 8.9671 0.00502 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Model 5

## Specifying Model 5

VAS_Model5 <- lm(VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b + vowel_ED_b, data = AcousticData)

## Model 5 Assumption Check

performance::check_model(VAS_Model5)

## Model 5 Summary

summary(VAS_Model5)

Call:
lm(formula = VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b + 
    vowel_ED_b, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-44.569 -13.697   5.187  16.183  40.204 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)  
(Intercept)   12.6419    23.9197   0.529   0.6006  
Hull_bVSD_25  -1.5667     1.3261  -1.181   0.2456  
Hull_bVSD_75   1.3137     2.4337   0.540   0.5928  
Hull_b         0.7124     0.8704   0.818   0.4188  
VSA_b          5.8159     2.9132   1.996   0.0539 .
vowel_ED_b     6.5644    14.6417   0.448   0.6568  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 23.99 on 34 degrees of freedom
Multiple R-squared:  0.2726,    Adjusted R-squared:  0.1657 
F-statistic: 2.549 on 5 and 34 DF,  p-value: 0.0461

## Model 4 and Model 5 Comparison

anova(VAS_Model4, VAS_Model5)

Analysis of Variance Table

Model 1: VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b
Model 2: VAS ~ Hull_bVSD_25 + Hull_bVSD_75 + Hull_b + VSA_b + vowel_ED_b
  Res.Df   RSS Df Sum of Sq     F Pr(>F)
1     35 19687                          
2     34 19572  1     115.7 0.201 0.6568

Final Model

## Specifying Final Model

VAS_Model_final <- lm(VAS ~ VSA_b, data = AcousticData)

## Final Model Assumption Check

performance::check_model(VAS_Model_final)

## Final Model Summary

summary(VAS_Model_final)

Call:
lm(formula = VAS ~ VSA_b, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-44.956 -15.943   6.754  17.153  43.062 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)   
(Intercept)   24.703      8.761   2.820  0.00760 **
VSA_b          6.163      1.798   3.427  0.00148 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 23.26 on 38 degrees of freedom
Multiple R-squared:  0.2361,    Adjusted R-squared:  0.216 
F-statistic: 11.74 on 1 and 38 DF,  p-value: 0.001482

confint(VAS_Model_final)

               2.5 %    97.5 %
(Intercept) 6.966936 42.438084
VSA_b       2.521887  9.803467

Research Q2: Relationship between OT and VAS

Model 1

# Specify Model

OT_VAS_model <- lm(transAcc ~ VAS*Etiology + VAS*Sex, data = AcousticData)

# Assumption Check

performance::check_model(OT_VAS_model)

# Model Results

summary(OT_VAS_model)

Call:
lm(formula = transAcc ~ VAS * Etiology + VAS * Sex, data = AcousticData)

Residuals:
    Min      1Q  Median      3Q     Max 
-14.910  -4.525  -1.280   5.529  16.932 

Coefficients:
                     Estimate Std. Error t value Pr(>|t|)    
(Intercept)        10.8993562  5.7092981   1.909   0.0659 .  
VAS                 0.8924319  0.1276625   6.991  9.1e-08 ***
EtiologyAtaxic      2.9854239  9.5149402   0.314   0.7559    
EtiologyHD          2.1009983  7.3948950   0.284   0.7783    
EtiologyPD         -2.4903757 10.1590855  -0.245   0.8080    
SexM                6.8939642  7.0796700   0.974   0.3380    
VAS:EtiologyAtaxic -0.0019088  0.1791048  -0.011   0.9916    
VAS:EtiologyHD      0.0007804  0.1453318   0.005   0.9958    
VAS:EtiologyPD      0.0320470  0.1732105   0.185   0.8545    
VAS:SexM           -0.1220661  0.1241230  -0.983   0.3333    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 8.488 on 30 degrees of freedom
Multiple R-squared:  0.9031,    Adjusted R-squared:  0.874 
F-statistic: 31.06 on 9 and 30 DF,  p-value: 8.181e-13

Final Linear Model

# Specify Final Model

OT_VAS_final <- lm(transAcc ~ VAS, data = AcousticData)

confint(OT_VAS_final)

                2.5 %     97.5 %
(Intercept) 8.1038688 19.3870846
VAS         0.7637645  0.9580856

# Model Results

summary(OT_VAS_final)

Call:
lm(formula = transAcc ~ VAS, data = AcousticData)

Residuals:
     Min       1Q   Median       3Q      Max 
-13.6882  -4.9316  -0.4408   4.9974  17.2110 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept) 13.74548    2.78681   4.932 1.64e-05 ***
VAS          0.86093    0.04799  17.938  < 2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 7.873 on 38 degrees of freedom
Multiple R-squared:  0.8944,    Adjusted R-squared:  0.8916 
F-statistic: 321.8 on 1 and 38 DF,  p-value: < 2.2e-16

Quadratic Linear Model

summary(OT_VAS_final_quadratic)

Call:
lm(formula = transAcc ~ VAS, data = AcousticData %>% dplyr::mutate(VAS = VAS^2))

Residuals:
    Min      1Q  Median      3Q     Max 
-18.729  -7.579   1.411   6.587  22.341 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept) 28.274672   2.894431   9.769 6.52e-12 ***
VAS          0.008956   0.000702  12.759 2.61e-15 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 10.54 on 38 degrees of freedom
Multiple R-squared:  0.8107,    Adjusted R-squared:  0.8058 
F-statistic: 162.8 on 1 and 38 DF,  p-value: 2.608e-15

Manuscript Tables

##Descriptives Table

gtData <- AcousticData %>%
  rbind(.,AcousticData %>%
          dplyr::mutate(Etiology = "All Etiologies")) %>%
  rbind(.,AcousticData %>%
          rbind(.,AcousticData %>%
          dplyr::mutate(Etiology = "All Etiologies")) %>%
          dplyr::mutate(Sex = "All")) %>%
  dplyr::mutate(Sex = as.factor(Sex),
                Etiology = as.factor(Etiology)) %>%
  dplyr::group_by(Sex, Etiology) %>%
  dplyr::summarize(VSA_mean = mean(VSA_b, na.rm =T), VSA_sd = sd(VSA_b, na.rm = T),
                   Disp_mean = mean(vowel_ED_b, na.rm =T), Disp_sd = sd(vowel_ED_b, na.rm =T),
                   Hull_mean = mean(Hull_b, na.rm =T), Hull_sd = sd(Hull_b, na.rm =T),
                   VSD25_mean = mean(Hull_bVSD_25, na.rm =T), VSD25_sd = sd(Hull_bVSD_25, na.rm =T),
                   VSD50_mean = mean(Hull_bVSD_50, na.rm =T), VSD50_sd = sd(Hull_bVSD_50, na.rm =T),
                   VSD75_mean = mean(Hull_bVSD_75, na.rm =T), VSD75_sd = sd(Hull_bVSD_75, na.rm =T),
                   VAS_mean = mean(VAS, na.rm =T), VAS_sd = sd(VAS, na.rm =T),
                   OT_mean = mean(transAcc, na.rm =T), OT_sd = sd(transAcc, na.rm =T)) %>%
  pivot_longer(cols = VSA_mean:OT_sd, names_to = "Measure",
               values_to = "Value") %>%
  dplyr::mutate(meanSD = ifelse(grepl("_mean",Measure),"M","sd"),
                Measure = gsub("_mean","",Measure),
                Measure = gsub("_sd","",Measure),
                Etiology = paste(Etiology,meanSD, sep = "_"),
                Sex = case_when(
                  Sex == "All" ~ "All Speakers",
                  Sex == "M" ~ "Male",
                  Sex == "F" ~ "Female"
                )) %>%
  dplyr::select(!meanSD) %>%
  pivot_wider(names_from = Etiology, values_from = "Value") %>%
  dplyr::filter(Measure != "VSD50")

`summarise()` has grouped output by 'Sex'. You can override using the `.groups` argument.

gtData %>%
  gt::gt(
    rowname_col = "Measure",
    groupname_col = "Sex",
  ) %>%
  fmt_number(
    columns = 'All Etiologies_M':PD_sd,
    decimals = 2
  ) %>%
  tab_spanner(
    label = "All Etiologies",
    columns = c('All Etiologies_M', 'All Etiologies_sd')
  ) %>%
    tab_spanner(
    label = "ALS",
    columns = c(ALS_M, ALS_sd)
  ) %>%
  tab_spanner(
    label = "PD",
    columns = c(PD_M, PD_sd)
  ) %>%
  tab_spanner(
    label = "HD",
    columns = c(HD_M, HD_sd)
  ) %>%
  tab_spanner(
    label = "Ataxic",
    columns = c(Ataxic_M, Ataxic_sd)
  ) %>%
  cols_label(
     'All Etiologies_M' = "M",
     'All Etiologies_sd' = "SD",
     ALS_M = "M",
     ALS_sd = "SD",
     PD_M = "M",
     PD_sd = "SD",
     HD_M = "M",
     HD_sd = "SD",
     Ataxic_M = "M",
     Ataxic_sd = "SD"
  ) %>%
  gtsave("DescriptivesTable.html", path = "Tables")

OT Model

sjPlot::tab_model(OT_Model1,OT_Model2,OT_Model3, OT_Model4, OT_Model5, OT_Model_final,
                  show.ci = F,
                  p.style = "stars",
                  file = "Tables/OT Models.html")

VAS Model

sjPlot::tab_model(VAS_Model1,VAS_Model2,VAS_Model3, VAS_Model4, VAS_Model5, VAS_Model_final,
                  show.ci = F,
                  p.style = "stars",
                  file = "Tables/VAS Models.html")

OT vs. VAS

sjPlot::tab_model(OT_VAS_model,OT_VAS_final,
                  show.ci = F,
                  show.reflvl = TRUE,
                  p.style = "stars",
                  file = "Tables/OT and VAS Comparison.html")

Manuscript Figures

Example Measures

formantColor <- "grey"
formantAlpha <- .95
lineColor <- "white"
lineAlpha <- .8

vowelData <- rio::import("Prepped Data/Vowel Data.csv") %>%
  dplyr::filter(Speaker == "AF8")

  Pitch_PRAAT <- list.files(path = paste("Prepped Data/Example Data/", sep = ""), 
                              pattern = ".Pitch", ignore.case = T) %>%
    paste("Prepped Data/Example Data/",., sep = "") %>%
    read.delim(., header = F) %>%
    dplyr::rename(Pitch = V1) %>%
    dplyr::mutate(Pitch = gsub("--undefined--",NA,Pitch),
                  Pitch = as.numeric(Pitch))
  
  Formants_PRAAT <- list.files(path = paste("Prepped Data/Example Data/", sep = ""), 
                              pattern = "_Formant", ignore.case = T) %>%
    paste("Prepped Data/Example Data/",., sep = "") %>%
    read.delim(., header = T) %>%
    dplyr::select(!c(nformants, B1.Hz., B2.Hz., B3.Hz., F4.Hz., B4.Hz., F5.Hz., B5.Hz.)) %>%
    dplyr::rename(Time_s = time.s.,
                  F1_Hz = F1.Hz.,
                  F2_Hz = F2.Hz.,
                  F3_Hz = F3.Hz.) %>%
    dplyr::mutate(F1_Hz = ifelse(F1_Hz == 0, NA, F1_Hz),
                  F2_Hz = ifelse(F2_Hz == 0, NA, F2_Hz),
                  F3_Hz = ifelse(F3_Hz == 0, NA, F3_Hz)) %>%
    dplyr::mutate(F1_Hz = as.numeric(F1_Hz),
                  F2_Hz = as.numeric(F2_Hz),
                  F3_Hz = suppressWarnings(as.numeric(F3_Hz)),
                  Time_ms = Time_s / 1000,
                  F1_kHz = F1_Hz / 1000,
                  F2_kHz = F2_Hz / 1000,
                  F3_kHz = F3_Hz / 1000) %>%
    dplyr::select(!Time_s) %>%
    dplyr::relocate(Time_ms, .before = F1_Hz) %>%
    cbind(.,Pitch_PRAAT)
  
  c <- 2
  while(c < NROW(Formants_PRAAT)){
    Formants_PRAAT$F1_Hz[c] <- ifelse(is.na(Formants_PRAAT$F1_Hz[c-1]) &&
                                        is.na(Formants_PRAAT$F1_Hz[c+1]),
                                      NA,
                                      Formants_PRAAT$F1_Hz[c])
    Formants_PRAAT$F2_Hz[c] <- ifelse(is.na(Formants_PRAAT$F2_Hz[c-1]) &&
                                        is.na(Formants_PRAAT$F2_Hz[c+1]),
                                      NA,
                                      Formants_PRAAT$F2_Hz[c])
    c <- c + 1
  }
  rm(c)
  
  Formants_PRAAT <- Formants_PRAAT %>%
    dplyr::filter(!is.na(Pitch)) %>%
    dplyr::mutate(F1_mad = (abs(F1_Hz - median(F1_Hz))/ mad(F1_Hz, constant = 1.4826)) > 2.5,
                  F2_mad = (abs(F2_Hz - median(F2_Hz))/ mad(F2_Hz, constant = 1.4826)) > 2.5) %>%
    dplyr::filter(F1_mad == FALSE & F2_mad == FALSE) %>%
    dplyr::mutate(mDist = mahalanobis(cbind(.$F1_Hz, .$F2_Hz),
                                      colMeans(cbind(.$F1_Hz, .$F2_Hz)),
                                      cov = cov(cbind(.$F1_Hz, .$F2_Hz))),
                  mDist_sd = abs(scale(mDist,center = T))) %>%
    dplyr::filter(mDist_sd < 2) %>%
    dplyr::select(!c(F1_mad, F2_mad, mDist, mDist_sd)) %>%
    dplyr::mutate(F1_z = scale(F1_Hz, center = TRUE),
                  F2_z = scale(F2_Hz, center = TRUE),
                  F3_z = scale(F3_Hz, center = TRUE),
                  F1_b = emuR::bark(F1_Hz),
                  F2_b = emuR::bark(F2_Hz),
                  F3_b = emuR::bark(F3_Hz))
  
  rm(Pitch_PRAAT)
  
  
## Corner Dispersion ----
  wedge <- vowelData %>%
    dplyr::group_by(Vowel) %>%
    dplyr::summarize(mean_F1 = mean(F1_tempMid),
              mean_F2 = mean(F2_tempMid),
              mean_F1_z = mean(F1_z_tempMid),
              mean_F2_z = mean(F2_z_tempMid),
              mean_F1_b = mean(F1_b_tempMid),
              mean_F2_b = mean(F2_b_tempMid)) %>%
    dplyr::filter(Vowel == "v")
    
  corner_dis <- vowelData %>%
    dplyr::filter(Vowel != "v") %>%
    dplyr::group_by(Vowel) %>%
    dplyr::summarize(mean_F1 = mean(F1_tempMid),
              mean_F2 = mean(F2_tempMid),
              mean_F1_z = mean(F1_z_tempMid),
              mean_F2_z = mean(F2_z_tempMid),
              mean_F1_b = mean(F1_b_tempMid),
              mean_F2_b = mean(F2_b_tempMid)) %>%
    dplyr::mutate(Vowel_ED = sqrt((mean_F1-wedge$mean_F1)^2 + (mean_F2-wedge$mean_F2)^2),
                  Vowel_ED_z = sqrt((mean_F1_z-wedge$mean_F1_z)^2 + (mean_F2_z-wedge$mean_F2_z)^2),
                  Vowel_ED_b = sqrt((mean_F1_b-wedge$mean_F1_b)^2 + (mean_F2_b-wedge$mean_F2_b)^2))

    
# Plot Corner Dispersion
      # Changing to IPA symbols
      corner_dis <- corner_dis %>%
        dplyr::mutate(Vowel = dplyr::case_when(
          Vowel == "ae" ~ "æ",
          TRUE ~ Vowel
        ))
      
      wedge <- wedge %>%
        dplyr::mutate(Vowel = case_when(
          Vowel == "v" ~ "ʌ",
          TRUE ~ Vowel
        ))
      
      CDplot <- ggplot(aes(x=F2_b,
                           y=F1_b),
                       data = Formants_PRAAT,
                       inherit.aes = FALSE) + 
      geom_point(shape = 21,
                 alpha = formantAlpha,
                 color = formantColor) + 
      geom_line(aes(x = mean_F2_b,
                    y = mean_F1_b),
                data = corner_dis %>%
                  dplyr::select(Vowel:mean_F2_b) %>%
                  dplyr::filter(Vowel == "i") %>%
                  rbind(.,wedge),
                color = lineColor,
                size = 1.5,
                alpha = lineAlpha) +
      geom_line(aes(x = mean_F2_b,
                    y = mean_F1_b),
                data = corner_dis %>%
                  dplyr::select(Vowel:mean_F2_b) %>%
                  dplyr::filter(Vowel == "a") %>%
                  rbind(.,wedge),
                color = lineColor,
                size = 1.5,
                alpha = lineAlpha) +
      geom_line(aes(x = mean_F2_b,
                    y = mean_F1_b),
                data = corner_dis %>%
                  dplyr::select(Vowel:mean_F2_b) %>%
                  dplyr::filter(Vowel == "æ") %>%
                  rbind(.,wedge),
                color = lineColor,
                size = 1.5,
                alpha = lineAlpha) +
      geom_line(aes(x = mean_F2_b,
                    y = mean_F1_b),
                data = corner_dis %>%
                  dplyr::select(Vowel:mean_F2_b) %>%
                  dplyr::filter(Vowel == "u") %>%
                  rbind(.,wedge),
                color = lineColor,
                size = 1.5,
                alpha = lineAlpha) +
      geom_point(aes(x = mean_F2_b,
                     y = mean_F1_b,
                     color = Vowel),
                 data = corner_dis %>%
                  dplyr::select(Vowel:mean_F2_b) %>%
                  rbind(.,wedge),
                 inherit.aes = FALSE,
                 size = 5) +
      scale_y_reverse() +
      scale_x_reverse() +
      theme_classic() + labs(title = paste("Corner Dispersion")) + xlab("F2 (bark)") + ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1) +
          scale_color_manual(values = c("a" = "#1AAD77",
                                        "æ" = "#1279B5",
                                        "i" = "#FFBF00",
                                        "u" = "#FD7853",
                                        "ʌ" = "#BF3178"))
    CDplot

    
      rm(corner_dis, wedge)
      
## Vowel Space Area ----
  VSA_coords <- vowelData %>%
    dplyr::filter(Vowel != "v") %>%
    dplyr::group_by(Vowel) %>%
    dplyr::summarize(mean_F1 = mean(F1_tempMid),
              mean_F2 = mean(F2_tempMid),
              mean_F1_z = mean(F1_z_tempMid),
              mean_F2_z = mean(F2_z_tempMid),
              mean_F1_b = mean(F1_b_tempMid),
              mean_F2_b = mean(F2_b_tempMid)) 
  
### Plotting VSA
    VSA_coords <- VSA_coords %>%
        dplyr::mutate(Vowel = case_when(
          Vowel == "ae" ~ "æ",
          TRUE ~ Vowel
        ))
    
    VSAplot <- ggplot(aes(x = F2_b,
                          y = F1_b),
                      data = Formants_PRAAT,
                      inherit.aes = FALSE) + 
      geom_point(shape = 21,
                 alpha = formantAlpha,
                 color = formantColor) + 
      geom_polygon(aes(x = mean_F2_b,
                       y = mean_F1_b),
                   data = VSA_coords,
                   alpha = lineAlpha,
                   color = lineColor,
                   fill=NA,
                   size = 1.5) +
      geom_point(aes(x = mean_F2_b,
                     y = mean_F1_b,
                     color = Vowel),
                 data = VSA_coords,
                 inherit.aes = FALSE,
                 size = 5) +
      scale_y_reverse() +
      scale_x_reverse() +
      guides(color = FALSE) +
      theme_classic() + labs(title = "VSA") + xlab("F2 (bark)") + ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1) +
                scale_color_manual(values = c("a" = "#1AAD77",
                                        "æ" = "#1279B5",
                                        "i" = "#FFBF00",
                                        "u" = "#FD7853"))

`guides(<scale> = FALSE)` is deprecated. Please use `guides(<scale> = "none")` instead.

    VSAplot

  
  rm(VSA_coords)
  
## Hull ----

### Plotting Hull
      convexCoords <- Formants_PRAAT %>%
        dplyr::select(F1_b, F2_b) %>%
        as.matrix() %>%
        grDevices::chull()
      convex <- Formants_PRAAT %>%
        slice(convexCoords)

      hullPlot <- ggplot(aes(F2_b, F1_b),
                         data = Formants_PRAAT) +
        geom_point(shape = 21,
                 alpha = formantAlpha,
                 color = formantColor) +
        geom_polygon(data = convex,
                     alpha = .5,
                     color = "#1279B5",
                     fill = NA,
                     size = 1.5) +
        scale_y_reverse() +
        scale_x_reverse() +
        theme_classic() + labs(title = "VSA Hull") + xlab("F2 (bark)") + ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1)
      hullPlot

    
  
## Vowel Space Density ----

## Bark Normalized Density ----
# selecting the bandwidth
H_hpi <- ks::Hpi(x = Formants_PRAAT[,c("F2_b","F1_b")], pilot = "samse", pre = "scale", binned = T)

# compute 2d kde
k <- kde(x = Formants_PRAAT[,c("F2_b","F1_b")],
         H = H_hpi,
         binned = T,
         gridsize = 250)

#density <- k[["estimate"]]

# Before we can plot the density estimate we need to melt it into long format
mat.melted <- data.table::melt(k$estimate)

The melt generic in data.table has been passed a matrix and will attempt to redirect to the relevant reshape2 method; please note that reshape2 is deprecated, and this redirection is now deprecated as well. To continue using melt methods from reshape2 while both libraries are attached, e.g. melt.list, you can prepend the namespace like reshape2::melt(k$estimate). In the next version, this warning will become an error.

names(mat.melted) <- c("x", "y", "density")

# We need to add two more colums to preserve the axes units
mat.melted$F2.b <- rep(k$eval.points[[1]], times = nrow(k$estimate))
mat.melted$F1.b <- rep(k$eval.points[[2]], each = nrow(k$estimate))
mat.melted$density <- scales::rescale(mat.melted$density, to = c(0, 1))

# VSD - 25
nVSD_25 <- mat.melted %>%
  dplyr::filter(density > .25) %>%
  dplyr::select(F2.b,F1.b, density) %>%
  dplyr::rename(Density = density)

convexCoords <- nVSD_25 %>%
  dplyr::select(F2.b, F1.b) %>%
  as.matrix() %>%
  #grDevices::xy.coords() %>%
  grDevices::chull()
nconvex_25 <- nVSD_25 %>%
  slice(convexCoords)

# VSD - 75
nVSD_75 <- mat.melted %>%
  dplyr::filter(density > .75) %>%
  dplyr::select(F2.b,F1.b, density) %>%
  dplyr::rename(Density = density)

convexCoords <- nVSD_75 %>%
  dplyr::select(F2.b, F1.b) %>%
  as.matrix() %>%
  grDevices::chull()
nconvex_75 <- nVSD_75 %>%
  slice(convexCoords)

# Plotting Z Normalized VSD 
    rf <- colorRampPalette(rev(RColorBrewer::brewer.pal(11, "Spectral")))
    r <- rf(32)
    
    plotData <- mat.melted %>%
                        dplyr::rename(Density = density) %>%
                        dplyr::mutate(VSDlabel = dplyr::case_when(
                          Density < .25 ~ "none",
                          Density > .25 && Density < .75 ~ "VSD25",
                          TRUE ~ "VSD75"
                        ))
geom.text.size <- 2
    VSDplot <- ggplot(data = plotData,
                      aes(x = F2.b,
                          y = F1.b,
                          fill = Density)) + 
      geom_tile() + 
      scale_fill_viridis_c() +
      #scale_fill_gradientn(colours = r) +
      scale_x_reverse(expand = c(0, 0), 
                      breaks = round(seq(min(mat.melted$F2.b), 
                                         max(mat.melted$F2.b)))) +
      scale_y_reverse(expand = c(0, 0),
                      breaks = round(seq(min(mat.melted$F1.b),
                                         max(mat.melted$F1.b)))) + 
      ylab("F1 (bark)") + xlab("F2 (bark)") +
      labs(title = "VSD Hull") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1) +
      geom_polygon(data = nconvex_25, alpha = lineAlpha, color = lineColor, size = 1.5, fill = NA, linetype = 1) +
      geom_polygon(data = nconvex_75, alpha = lineAlpha, color = lineColor, size = 1.5, fill = NA, linetype = 2) +
      geom_labelcurve(data = data.frame(x = 10.382452, xend = 8.313515, y = 2.874967, yend = 4.127919, Density = 1), 
                 aes(x, y, xend = xend, yend = yend),
                 hjust = 0.5,
                 vjust = 0,
                 curvature = 0,
                 size = geom.text.size,
                 label = "VSD 25",
                 text_only = T,
                 color = "black",
                 fill = "white",
                 rich = TRUE) +
        geom_labelcurve(data = data.frame(x = 14.92512, xend = 12.04660, y = 3.95, yend = 6, Density = 1), 
                 aes(x, y, xend = xend, yend = yend),
                 hjust = 0.3, 
                 vjust = 0,
                 curvature = 0,
                 size = geom.text.size,
                 fill = "white",
                 label = "VSD 75",
                 text_only = T,
                 color = "black")
     VSDplot

# Combined Plot

measuresPlot <- VSAplot + CDplot + hullPlot + VSDplot +
      patchwork::plot_layout(guides = 'collect',
                             ncol = 2)

ggsave(filename = "Plots/Measures.png",
       plot = measuresPlot,
       height = 7,
       width = 8,
       scale = .8)

Filtering Process

formantAlpha <- .20
myPal <- c("#1279B5","#2D2D37")

Pitch_PRAAT <- list.files(path = paste("Prepped Data/Example Data/", sep = ""), 
                              pattern = ".Pitch", ignore.case = T) %>%
    paste("Prepped Data/Example Data/",., sep = "") %>%
    read.delim(., header = F) %>%
    dplyr::rename(Pitch = V1) %>%
    dplyr::mutate(Pitch = gsub("--undefined--",NA,Pitch),
                  Pitch = as.numeric(Pitch))

Formants_PRAAT <- list.files(path = paste("Prepped Data/Example Data/", sep = ""), 
                              pattern = "_Formant", ignore.case = T) %>%
    paste("Prepped Data/Example Data/",., sep = "") %>%
    read.delim(., header = T) %>%
    dplyr::select(!c(nformants, B1.Hz., B2.Hz., B3.Hz., F4.Hz., B4.Hz., F5.Hz., B5.Hz.)) %>%
    dplyr::rename(Time_s = time.s.,
                  F1_Hz = F1.Hz.,
                  F2_Hz = F2.Hz.,
                  F3_Hz = F3.Hz.) %>%
    dplyr::mutate(F1_Hz = ifelse(F1_Hz == 0, NA, F1_Hz),
                  F2_Hz = ifelse(F2_Hz == 0, NA, F2_Hz),
                  F3_Hz = ifelse(F3_Hz == 0, NA, F3_Hz)) %>%
    dplyr::mutate(F1_Hz = as.numeric(F1_Hz),
                  F2_Hz = as.numeric(F2_Hz),
                  F3_Hz = suppressWarnings(as.numeric(F3_Hz)),
                  Time_ms = Time_s / 1000,
                  F1_kHz = F1_Hz / 1000,
                  F2_kHz = F2_Hz / 1000,
                  F3_kHz = F3_Hz / 1000,
                  F1_b = emuR::bark(F1_Hz),
                  F2_b = emuR::bark(F2_Hz),
                  F3_b = emuR::bark(F3_Hz)) %>%
    dplyr::select(!Time_s) %>%
    dplyr::relocate(Time_ms, .before = F1_Hz) %>%
    cbind(.,Pitch_PRAAT)
  
  c <- 2
  while(c < NROW(Formants_PRAAT)){
    Formants_PRAAT$F1_Hz[c] <- ifelse(is.na(Formants_PRAAT$F1_Hz[c-1]) &&
                                        is.na(Formants_PRAAT$F1_Hz[c+1]),
                                      NA,
                                      Formants_PRAAT$F1_Hz[c])
    Formants_PRAAT$F2_Hz[c] <- ifelse(is.na(Formants_PRAAT$F2_Hz[c-1]) &&
                                        is.na(Formants_PRAAT$F2_Hz[c+1]),
                                      NA,
                                      Formants_PRAAT$F2_Hz[c])
    c <- c + 1
  }
  rm(c)
  
  # Raw Formants ----
  f1 <- ggplot(aes(x=F2_b,
                   y=F1_b),
               data = Formants_PRAAT) + 
      geom_point(shape = 21, color = myPal[2]) +
      scale_y_reverse() +
      scale_x_reverse() +
      scale_color_manual(values = myPal) +
      theme_classic() + labs(title = paste("Raw Formant Values")) + xlab("F2 (bark)") + ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1,
            legend.title = element_blank(),
            legend.text = element_text(size=12))
    
# Step #1: Voiced Segments ----
    plotData <- Formants_PRAAT %>%
                   dplyr::mutate(isOutlier = case_when(
                     is.na(Pitch) ~ "Removed",
                     TRUE ~ "Retained"
                   ))
    f2 <- ggplot(data = plotData,
                 aes(x = F2_b,
                     y = F1_b,
                     color = isOutlier)) + 
      geom_point(shape = 21, data = plotData %>%
                   dplyr::filter(isOutlier == "Removed")) +
      geom_point(shape = 21, data = plotData %>%
             dplyr::filter(isOutlier == "Retained")) +
      scale_y_reverse() +
      scale_x_reverse() +
      scale_color_manual(values = myPal) +
      theme_classic() + labs(title = paste("Step #1:\nVoiced Segments")) +
      xlab("F2 (bark)") +
      ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1,
            legend.title = element_blank(),
            legend.text = element_text(size=12))
    
# Step 2: MAD ----
    plotData <- Formants_PRAAT %>%
      dplyr::filter(!is.na(Pitch)) %>%
      dplyr::mutate(F1_mad = (abs(F1_Hz - median(F1_Hz))/ mad(F1_Hz, constant = 1.4826)) > 2.5,
                    F2_mad = (abs(F2_Hz - median(F2_Hz))/ mad(F2_Hz, constant = 1.4826)) > 2.5,
                    isOutlier = case_when(
                      F1_mad == TRUE | F2_mad == TRUE ~ "Removed",
                      TRUE ~ "Retained"
               ))
    
    f3 <- ggplot(data = plotData,
                 aes(x = F2_b,
                     y = F1_b,
                     color = isOutlier)) + 
      geom_point(shape = 21, data = plotData %>%
                   dplyr::filter(isOutlier == "Removed")) +
      geom_point(shape = 21, data = plotData %>%
             dplyr::filter(isOutlier == "Retained")) +
      scale_y_reverse() +
      scale_x_reverse() +
      scale_color_manual(values = myPal) +
      theme_classic() +
      labs(title = paste("Step #2:\nMedian Absolute Deviation")) +
      xlab("F2 (bark)") +
      ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1,
            legend.title = element_blank(),
            legend.text = element_text(size=12))
    
# Step 3: Mahalanhobis Distance ----
  plotData <- Formants_PRAAT %>%
      dplyr::filter(!is.na(Pitch)) %>%
      dplyr::mutate(F1_mad = (abs(F1_Hz - median(F1_Hz))/ mad(F1_Hz, constant = 1.4826)) > 2.5,
                    F2_mad = (abs(F2_Hz - median(F2_Hz))/ mad(F2_Hz, constant = 1.4826)) > 2.5) %>%
      dplyr::filter(F1_mad == FALSE & F2_mad == FALSE) %>%
      dplyr::mutate(mDist = mahalanobis(cbind(.$F1_Hz, .$F2_Hz),
                                      colMeans(cbind(.$F1_Hz, .$F2_Hz)),
                                      cov = cov(cbind(.$F1_Hz, .$F2_Hz))),
                  mDist_sd = abs(scale(mDist,center = T)),
                  isOutlier = case_when(
                    mDist_sd < 2 ~ "Retained",
                    TRUE ~ "Removed"
                  ))
    
    f4 <- ggplot(data = plotData,
                 aes(x = F2_b,
                     y = F1_b,
                     color = isOutlier)) + 
      geom_point(shape = 21, data = plotData %>%
                   dplyr::filter(isOutlier == "Removed")) +
      geom_point(shape = 21, data = plotData %>%
             dplyr::filter(isOutlier == "Retained")) +
      scale_y_reverse() +
      scale_x_reverse() +
      scale_color_manual(values = myPal) +
      theme_classic() + labs(title = paste("Step #3:\nMahalanobis Distance")) + xlab("F2 (bark)") + ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1,
            legend.title = element_blank(),
            legend.text = element_text(size=12))
    
# Final Formants ----
    plotData <- Formants_PRAAT %>%
    dplyr::filter(!is.na(Pitch)) %>%
    dplyr::mutate(F1_mad = (abs(F1_Hz - median(F1_Hz))/ mad(F1_Hz, constant = 1.4826)) > 2.5,
                  F2_mad = (abs(F2_Hz - median(F2_Hz))/ mad(F2_Hz, constant = 1.4826)) > 2.5) %>%
    dplyr::filter(F1_mad == FALSE & F2_mad == FALSE) %>%
    dplyr::mutate(mDist = mahalanobis(cbind(.$F1_Hz, .$F2_Hz),
                                      colMeans(cbind(.$F1_Hz, .$F2_Hz)),
                                      cov = cov(cbind(.$F1_Hz, .$F2_Hz))),
                  mDist_sd = abs(scale(mDist,center = T))) %>%
    dplyr::filter(mDist_sd < 2)
    
    f5 <- ggplot(aes(x=F2_b,
                     y=F1_b),
                       data = plotData,
                       inherit.aes = FALSE) + 
      geom_point(shape = 21, color = myPal[2]) + 
      scale_y_reverse() +
      scale_x_reverse() +
      theme_classic() + labs(title = paste("Final Formant Values")) + xlab("F2 (bark)") + ylab("F1 (bark)") +
      theme(plot.title = element_text(hjust = 0.5),
            aspect.ratio = 1, legend.title = element_blank())
    
# Comibing plots
    filteredPlot <- f1 + f2 + f3 + f4 + f5 + patchwork::guide_area() +
      patchwork::plot_layout(guides = 'collect',
                         ncol = 3)
    filteredPlot
    
    ggsave(plot = filteredPlot, "Plots/Filtered Formants.png",
           height = 6,
           width = 8,
           units = "in",
           scale = .9)
  

OT vs. VAS

plotData_Int <- AcousticData %>%
  dplyr::filter(!grepl("_rel", Speaker)) %>%
  dplyr::group_by(Speaker) %>%
  dplyr::mutate(segMin = base::min(VAS, transAcc),
                segMax = base::max(VAS, transAcc),
                ratingAvg = mean(VAS, transAcc, na.rm = T),
                Speaker = as.factor(Speaker),
                Etiology = as.factor(Etiology)) %>%
  arrange(segMax)

my_pal <- c("#f26430", "#272D2D","#256eff")
# With a bit more style
plot_Int <- ggplot(plotData_Int) +
  geom_segment(aes(x = fct_inorder(Speaker),
                   xend = Speaker,
                   y = segMin,
                   yend = segMax,
                   color = Etiology)) +
  geom_point(aes(x = Speaker,
                 y = VAS,
                 color = Etiology),
             #color = my_pal[1],
             size = 3,
             shape = 19) +
  geom_point(aes(x = Speaker,
                 y = transAcc,
                 color = Etiology),
             #color = my_pal[2],
             size = 3,
             shape = 15) +
  coord_flip()+
  theme_classic() +
  theme(
    legend.position = "none",
    panel.border = element_blank(),
  ) +
  xlab("") +
  ylab("Speech Intelligibility") +
  ggtitle("Speech Intelligibility") +
  ylim(c(0,100))
plot_Int

myPal <- c("#1AAD77", "#1279B5", "#FFBF00", "#FD7853", "#BF3178")
myShapes <- c(16, 18, 17, 15)

scatter1 <- ggplot(plotData_Int,
                  aes(x = VAS,
                      y = transAcc,
                      color = Etiology,
                      shape = Etiology,
                      linetype = Etiology)) +
  geom_point() +
  #geom_smooth(method = "lm", se = F) +
  geom_abline(intercept = 0, slope = 1) +
  coord_cartesian(xlim = c(0,100), ylim = c(0,100)) +
  labs(x = "Intelligibility (VAS)", y = "Intelligibility (OT)") +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myShapes) +
  theme_classic() +
  theme(aspect.ratio=1,
        legend.position="right")

scatter2 <- ggplot(plotData_Int,
                  aes(x = VAS,
                      y = transAcc,
                      color = Etiology,
                      shape = Etiology,
                      linetype = Etiology)) +
  #geom_point() +
  geom_smooth(method = "lm", se = F) +
  geom_abline(intercept = 0, slope = 1) +
  coord_cartesian(xlim = c(0,100), ylim = c(0,100)) +
  labs(x = "Intelligibility (VAS)", y = "Intelligibility (OT)") +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myShapes) +
  theme_classic() +
  theme(aspect.ratio=1,
        legend.position="right")

combinedScatter <- ggarrange(scatter1,
                             scatter2,
                             common.legend = F,
                             ncol = 2,
                             nrow = 1)
combinedScatter

ggsave(filename = "Plots/OT and VAS Scatterplot.png",
       plot = combinedScatter,
       height = 2,
       width = 6,
       units = "in",
       scale = 1)

rm(scatter1, scatter2, combinedScatter)

Model Scatterplot

modelFigureData <- AcousticData %>%
  dplyr::filter(!grepl("_rel",Speaker)) %>%
  dplyr::select(Speaker, Etiology, Sex, VSA_b, vowel_ED_b, Hull_b, Hull_bVSD_25, Hull_bVSD_75, VAS, transAcc) %>%
  dplyr::mutate(Speaker = as.factor(Speaker),
                Etiology = as.factor(Etiology),
                Sex = as.factor(Sex)) %>%
  tidyr::pivot_longer(cols = VAS:transAcc, names_to = "IntType", values_to = "Int") %>%
  dplyr::mutate(IntType = case_when(
    IntType == "transAcc" ~ "OT",
    TRUE ~ "VAS"
  ),
                IntType = as.factor(IntType))

ylabel <- "Intelligibility"
myPal <- c("#2D2D37", "#1279B5")
myPalShape <- c(19, 1)

VSA <- modelFigureData %>%
  ggplot() +
  aes(x = VSA_b,
      y = Int,
      color = IntType,
      shape = IntType,
      linetype = IntType) +
  geom_point() +
  geom_smooth(method = "lm", se = T, fill = "light grey") +
  xlab(expression("VSA (Bark"^2*")")) +
  ylab(ylabel) +
  coord_cartesian(ylim = c(0,100)) +
  theme_classic() +
  theme(plot.title = element_text(hjust = 0.5, face = "bold"),
        aspect.ratio=1) +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myPalShape) +
  labs(color="Intelligibility Type",
       shape = "Intelligibility Type",
       linetype = "Intelligibility Type")

disp <- modelFigureData %>%
  ggplot() +
  aes(x = vowel_ED_b,
      y = Int,
      color = IntType,
      shape = IntType,
      linetype = IntType) +
  geom_point() +
  geom_smooth(method = "lm", se = T, fill = "light grey") +
  xlab("Corner Dispersion (Bark)") +
  ylab(ylabel) +
  coord_cartesian(ylim = c(0,100)) +
  theme_classic() +
  theme(aspect.ratio=1) +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myPalShape) +
  labs(color="Intelligibility Type",
       shape = "Intelligibility Type",
       linetype = "Intelligibility Type")

Hull <- modelFigureData %>%
  ggplot() +
  aes(x = Hull_b,
      y = Int,
      color = IntType,
      shape = IntType,
      linetype = IntType) +
  geom_point() +
  geom_smooth(method = "lm", se = T, fill = "light grey") +
  xlab(expression("VSA"[Hull]*" (Bark"^2*")")) +
  ylab(ylabel) +
  coord_cartesian(ylim = c(0,100)) +
  theme_classic() +
  theme(aspect.ratio=1) + theme(legend.position = "none") +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myPalShape) +
  labs(color="Intelligibility Type",
       shape = "Intelligibility Type",
       linetype = "Intelligibility Type")

vsd25 <- modelFigureData %>%
  ggplot() +
  aes(x = Hull_bVSD_25,
      y = Int,
      color = IntType,
      shape = IntType,
      linetype = IntType) +
  geom_point() +
  geom_smooth(method = "lm", se = T, fill = "light grey") +
  xlab(expression("VSD"[25]*" (Bark"^2*")")) +
  ylab(ylabel) +
  coord_cartesian(ylim = c(0,100)) +
  theme_classic() +
  theme(aspect.ratio=1) + theme(legend.position = "none") +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myPalShape) +
  labs(color="Intelligibility Type",
       shape = "Intelligibility Type",
       linetype = "Intelligibility Type")

vsd75 <- modelFigureData %>%
  ggplot() +
  aes(x = Hull_bVSD_75,
      y = Int,
      color = IntType,
      shape = IntType,
      linetype = IntType) +
  geom_point() +
  geom_smooth(method = "lm", se = T, fill = "light grey") +
  xlab(expression("VSD"[75]*" (Bark"^2*")")) +
  ylab(ylabel) +
  coord_cartesian(ylim = c(0,100)) +
  theme_classic() +
  theme(aspect.ratio=1) + theme(legend.position = "none") +
  scale_color_manual(values = myPal) +
  scale_shape_manual(values = myPalShape) +
  labs(color="Intelligibility Type",
       shape = "Intelligibility Type",
       linetype = "Intelligibility Type")

# Creating OT Scatterplot Figure

scatter <- VSA + disp + patchwork::guide_area() + Hull + vsd25 + vsd75 +
  patchwork::plot_layout(guides = 'collect',
                         ncol = 3)
scatter 

ggsave("Plots/ModelFigure.png", scatter,
       height = 4,
       width = 6,
       units = "in",
       scale = 1.1)

Listener Demographic Information

ListenerDemo <- Listeners %>%
  furniture::table1(age, gender, race, ethnicity)

ListenerDemo

Speaker Demographics

SpeakerDemo <- AcousticData %>%
  dplyr::select(c(Speaker, Sex, Etiology))

Ages <- rio::import("Prepped Data/Speaker Ages.xlsx")

SpeakerDemo <- full_join(SpeakerDemo, Ages, by = "Speaker")

SpeakerDemoInfo <- SpeakerDemo %>%
  furniture::table1(Sex, Etiology, Age, na.rm = F)

SpeakerDemoInfo

SpeakerDemo %>%
  dplyr::summarize(mean_age = mean(Age, na.rm = T), age_sd = sd(Age, na.rm = T), age_range = range(Age, na.rm = T))